Slow simulation

[Arthur92710]

Why are the simulations so slow?
My pc is pretty good 3.1Ghz dual core, 2gigs ram, windows xp.
It should be able to handle multisim but it works so slow.
Multisim also lags really bad very often. I have had some problems with programs that dont support dual core cpus but an update has fixed those problems.
What can I do?

[lacy]

This is what I know. The more complex the circuit the longer it takes to simulate. Also, the higher the frequency in the circuit the longer it takes to simulate. A fast computer does help, but it still takes the Spice engine time to crunch all the numbers for the circuit.  This is why it runs slow in some cases. Very simple circuits can fly, but a digital computer simulation is going to crawl. This is just the nature of the beast and there are very limited options to speed it up.

One option that I told another user is to make sure you defrag your harddrive. This seemed to help his slowdwon and it could possibly help in yours. This and adjusting TMAX under the SIMULATE>INTERACTIVE SIMULATION SETTINGS to 1e-003 is about all I know to increase the speed. Adjusting TMAX though is tricky as sometimes after doing this your get errors popping up saying "Timestep too small". 

I hope some of this may be of use and I was able to help you understand the reason behind the slowness.

[Arthur92710]

My curcuit is not very complex(I think) Its just a digital clock, and it I want to test it making TMAX so small would not work for me. Is there anything else that I can do?

[lacy]

There is not much that can be done. It will simulate, but it will be agonizingly slow. I don't know anything about how you designed it, but from the schematics I have seen in various literature there are a a number of IC's to do the frequency division necessary,  the counters, plus the IC's used to drive the displays, etc. This to me is a fairly large design. When I said simple circuits I was talking more in the line of a Voltage Source and maybe a few components like resistors and capacitors. Power Supply circuits simulate fairly quickly because they only use a handful of parts. So basically, the more components the slower the simulation as there are more things the simulator has to do.

As far as what to do, that is a question that is tough to answer. If you are using a 60Hz reference for the time base you could try increasing this to like 1Khz. Theoretically this will increase the speed of the time division portion of the circuit and may help as instead of taking 1 second to display one second on the displays it would do that in a fraction of the time. Sometimes this works, but usually I get a slow down when the frequency of the circuit increases (this is my personal observation and it may just happen to me).  You could give this a shot and see what might happen. The worse thing that could happen is that it won't work.

You could try 2 other options if TMAX isn't doing it for you. Goto SIMULATE>INTERACTIVE SIMULATION SETTINGS>ANALYSIS OPTION. Click the circle on Use Custom Settings and then on Cutomize. On the first page look for RELTOL  click the square to activate it and increase (or is it reduce?) its value from .001 to .01. 

Message Edited by lacy on 04-01-2008 08:07 PM

[lacy]

Here is the second option that didn't get posted because I accidently hit the Submit button before I was finished.

In this same Customize menu goto the TRANSIENT Tab and look down to the bottom of the list for METHOD and click the square to activate it. In the drop down box try selecting either trapezodial or gear. Try one and then the other to see if that helps with the speed and use the one that gives the faster results.

When I do time based circuits I usually split them up into sections. I did a timer circuit that used a dsiplay to show the numbers, but instead of having my oscillator (555 timer) on the same schematic I eliminated it and replaced it with a Clock Source as I knew that including the timer would slow down the simulator when trying to design the rest of the circuit. I tested the counter & display circuits on one schematic and the timer on the other. I then combined them (manually although I could have used a hierarchy block) and then tested them together. This way the simulator could be a little faster when troubleshooting my design errors. When I combined them, I did get a significant slow down as I expected.

This is about all know on how to increase the simualtion speed without causing errors to start popping up. If there are any other methods then I just don't know about them.

[Arthur92710]

Thanks, Ill try that out.

[Lartlace]

Multisim V9 Too Slow !!!  Original files are fast.  Later files are slow.  Even moving parts can take seconds. Possibly the latest circuit files are put on the disk much further away.  I don't know how to solve this yet.

Lartlace

[lacy]

Didin't the Defrag help?  In your other thread you said it helped so I assumed it did. If newly saved files are still slow then I don't have an answer to that. That will have to be addressed by NI since I don't have Version 9 to run tests on. I was hoping that something of a general nature was causing this, but it is probably something deeper in the software that is at issue here that I just don't have a frame of reference in order to make a diagnoses. Sorry, I did the best with what I knew at the time I knew it.

Message Edited by lacy on 04-02-2008 06:09 PM

[Arthur92710]

Replacing the power supply with a Clock source segnifigantly increases simulation speed. but it lag at random parts.

[lacy]

I wish there was a magic solution that we both could use to make the simulator run faster, but unlike Staples we don't have an "easy" button (laughing). I know it can be fustrating waiting 3 real world minutes for the simulator to do 10 seconds of circuit time as I have done that before.

This situation is just the nature of using the time-domain simulation (transient) mode of Multisim and speeding it up is only going to be possible up to a certain point. This is why I break my work up into bite sized pieces. Once all the errors are ironed out in each portion. I then tie them together and simulate them as one unit knowing ahead of time that it is going to be slower as a whole than it was individually.

I wih there was more I could tell you, but I have reached the limit of what I know can be done. If there are any other suggestions from other users or NI I am sure we would both be interested in hearing it.